MCP for Hotels: Is It Safe to Connect Your PMS to AI?
MCP is an open standard that lets AI assistants like Claude and ChatGPT read data from business software. For hotels, it means asking your PMS questions in your own AI app, with read-only access limited to the approving staff member’s role.
By the Frontdesko team • Last updated 11 November 2026
What is MCP, in plain language?
MCP, the Model Context Protocol, is an open standard that lets an AI assistant connect to the software where your data lives. Anthropic introduced it on 25 November 2024 as “a new standard for connecting AI assistants to the systems where data lives” (Anthropic announcement). Think of it as one plug shape: a PMS that speaks it can be connected to any AI app that also speaks it, without a custom integration for each.
For a hotel, the PMS is the system that holds arrivals, rooms, folios and revenue. An MCP connector is the permissioned door through which an AI app, such as Claude or ChatGPT, can read those records when you ask.
Why does an MCP-ready PMS matter for a small hotel?
- You ask where you already work. The owner asks in the AI app she uses for everything else, instead of opening report screens.
- No export-and-paste. Nobody downloads a spreadsheet to feed an AI. The assistant reads the live record, so a question at 18:30 is answered with the 18:30 numbers.
- One connector, several AI apps. It works with Claude, ChatGPT and other apps that support remote connectors, so you are not tied to one assistant.
- Permissions travel with it. The AI gets the approver’s role, not a master key.
- It is a small-team shortcut. A 12-room property with one manager has no analyst. A question like “which nights next month are below half full?” becomes a ten-second task. See 29 questions to ask.
Frontdesko’s private connector exposes 19 read-only lookups for one hotel. It is separate from the public connector on the MCP page, which answers questions about Frontdesko itself and has no hotel data. For the buyer view of AI features, read what an AI-ready hotel PMS means; the in-app version is the AI assistant.
Is it safe to connect my PMS to ChatGPT or Claude?
It is as safe as the controls around it, so here they are, plainly. The connection is read-only and runs as a named staff member, with that role’s limits. What the AI reads appears in your conversation with the AI company and is handled under that company’s terms, so connect only accounts you trust and ask only what you need. Names, phone numbers and balances can appear in answers.
Setup steps are in how to connect your PMS to ChatGPT or Claude; roles and codes are covered in staff roles and sign-in codes.
What does it look like when the controls kick in?
Two illustrative scenarios at an invented guesthouse. No real customers, people or events.
A manager asks for something only an admin can see
The manager approved the connection and asks Claude to sweep the hotel for double-booked rooms and overdue check-outs. That health check is admin only, so the lookup is refused, and Claude tells her which access she needs and that her hotel admin grants it. She can still ask about arrivals, revenue and payments, because a manager’s role allows them. Nothing was changed, and the lookups are in the log.
A lost laptop
The owner leaves a laptop with Claude signed in on a train. She emails Frontdesko support and asks them to revoke her AI connection. The next request from that laptop is refused. Had she not noticed, the connection would also have lapsed after 60 days without use. Because it was read-only, the worst case was someone reading what her role could read; no one could cancel a booking or take a payment. She approves again from a new device.
What should you ask any PMS vendor about AI access?
A checklist you can copy into an email to any vendor, ours included. Good answers are specific and say what is and is not possible.
- Is the AI access read-only, and can you show me where the write permission is, if any?
- Who can authorise a connection, and does it use my staff sign-in with a second factor?
- Does the AI inherit the approving person’s role, or does it get a separate, broader key?
- Can I see a log of every question the AI asked, by person and time?
- Can I switch AI access off for my whole property, and block single lookups, without a support ticket? If not, how fast is support?
- How do I revoke one connection, and how quickly does revocation take effect?
- How long does an unused connection stay valid?
- Do I paste a password or API key into the AI app, or do I approve on your page?
- Which data fields can appear in answers: guest names, phones, balances?
- Does it work with more than one AI app, or only one vendor’s assistant?
- Is access included in my plan, and is it priced separately?
- What does the AI company do with what it reads, and where is that written down?
Frontdesko’s answers are in the table above. Where there is no self-service screen yet (disconnecting, the off switch, the block list), we say so rather than imply it exists. For the buyer-side framing, see AI-ready hotel PMS.
How do I try it?
Start with the free PMS and a demo property, ask support to confirm the connector for your plan, then approve it from Claude or ChatGPT. Begin with a safe question such as “Who arrives today?” and compare the answer with the dashboard. Pricing for the PMS and paid add-ons is on the pricing page.
Frequently Asked Questions
What is an MCP server for a hotel?
It is the connector that lets an AI app such as Claude or ChatGPT read a hotel’s PMS records through a permissioned, standard interface. Frontdesko’s version is read-only, works for one hotel per connection, and runs with the approving staff member’s role.
Is it safe to connect my PMS to ChatGPT?
The connection is read-only, approved by an admin or manager after the normal staff login with an emailed code, limited to that person’s role, and every lookup is logged. What the AI reads shows in your chat with the AI company, so connect only trusted accounts.
Can the AI edit bookings or take payments through MCP?
No. The one-click connection has the read scope only, so it cannot create or change reservations, post payments or add charges. Write access is not available yet.
How do I disconnect or revoke access?
An admin or manager opens Settings, AI connectors in the PMS and clicks Revoke next to the connection. It takes effect on the next request. Admins can also switch AI access off for the whole hotel or untick individual tools there.
How long does the connection last?
Access passes last one hour and renew automatically while the connection is in use. The renewal window slides for 60 days from last use. After that, the admin or manager approves again from the AI app.
Who can see what the AI asked?
Frontdesko logs every lookup with the hotel, person, lookup name, result, duration and AI app. Your own conversation history with the AI company shows the questions and answers on its side.
Does it work with Claude and ChatGPT?
Yes, both can add Frontdesko as a custom connector, as can other apps that support remote connectors. Menu names and plan availability differ by app, so check your AI plan.
Which Frontdesko plan includes the connector?
This page does not state a plan requirement for the connector because it is not confirmed. The pricing page lists what each plan includes, and Frontdesko support can confirm for your hotel.
Ask Your Own Hotel a Question
Try Frontdesko’s free PMS with a demo property, then see how the AI connector fits your hotel.
Related Guides
Keep reading
29 Questions to Ask Your Hotel’s AI
Copy-ready prompts mapped to the lookup and role.
Connect Your PMS to ChatGPT or Claude
Step-by-step setup and limits.
Public MCP Server
The developer page for the public connector.
AI-Ready Hotel PMS
What to ask a vendor about AI readiness.
Roles and Sign-In Codes
Who can approve and what each role sees.
Multi-Property Reporting
Reporting across several properties.